I bet OTC is all over this article... :-P

Talk about timing. OTC here's something I'm sure caught your eye... http://www.theregister.co.uk/2009/09/17/idc_ca_mainframe_surveys/ I'm sorry, but I still would rather look at building distributed Unix boxes instead of investing the bulk of the money on a large centralized machine that is probably overkill for my client's needs. The report from IDS costs $20,000.00 USD. You can bet one of two things. 1) They don't expect to sell a lot of copies of the report. 2) Their target audience is the mainframe shops of the world. (They're used to overpaying for *everything* ) :-) Granted I'm not looking at projects where they want extremely large DW and supporting 1000's of concurrent users. Even there, the apps tend to be web based, so that's distributed, and then the data sources could be distributed as well. I'm sorry, I'm not convinced that you can't build out a network of COTS (Off the shelf kit) that are more power efficient and can handle the load better. It would be nice if we had a couple of TPC-C benchmarks out there (Or its equivalent) so that we can gauge how large of a system we would need to build out based on our application. So if you need 20,000 tpm/s capacity during the full day, you know that you'll have to build a system that could achieve that under a minimum load, so you have room to grow. If I know I only need a machine that does less than 1,000 tpm concurrently, I would do a much smaller system. So not to bitch, but having some sort of standardized benchmark that we could gauge systems against would be a good thing. Ok without it, we can still do this by looking at the machine size for a given benchmark result and *ass*u*me* that IDS would be roughly the same results or within an order of magnitude of performance. _________________________________________________________________ Ready for Fall shows? Use Bing to find helpful ratings and reviews on digital tv's. http://www.bing.com/shopping/search?q=digital+tv's&form=MSHNCB&publ=WLHMTAG&crea=TEXT_MSHNCB_Vertical_Shopping_DigitalTVs_1x1

Ian Michael Gumby wrote: > So not to bitch, but having some sort of standardized benchmark that we > could gauge systems against would be a good thing. This is a point where generically speaking we could all agree... but... > Ok without it, we can still do this by looking at the machine size for a > given benchmark result and *ass*u*me* that IDS would be roughly the same > results or within an order of magnitude of performance. ouch... you have a technical background... do you imagine that the published results are near the reality? Using your sort of speach, I could say that I know of a customer who picked some old (roughly the same age as their current machines) numbers and calculated that if my current machines were able to do that, and the new ones are able to do this (new benchmark numbers) than my new machines sizing will be "this". They completley ignored the ammount of memory and disks the machines had. They also ignored the tweaks that are made for benchmark purposes that you cannot do in real systems etc. Hell, they even missed that the CPU versions were not exactly the ones they were using (they were older, and slower). My point (which no doubtly you'll spend time to destroy) is that this may look good for marketing and market positions, but plase don't try to sell us the idea that it makes sense technically... The above is what happen when "cientists" try to transpose the benchmark numbers into reality... Regards.